Inspiration

We wanted to make learning for kids feel natural, fun, and safe — like chatting with a friendly teacher or playing pretend with a friend. Many children learn faster by talking, not just reading or writing. So, we thought: What if there was a voice AI that kids could talk to in real time — one that teaches, plays roleplay games, and explains things simply? That’s how GenieTalk was born.

What it does

GenieTalk is a real-time AI voice tutor for children aged 5–12. It works in two main modes:

Tutor Mode — teaches basic subjects like math, grammar, and science in simple, kid-friendly language.

Roleplay Mode — acts as a doctor, shopkeeper, or teacher, letting kids practice real-life conversations.

Children can talk directly to GenieTalk using their microphone, and the AI answers with a warm, friendly voice. A transcript is shown for reading practice. The interface is bright, clean, and easy for young kids to use.

How we built it

Frontend: HTML, CSS, JavaScript (mobile-friendly and responsive).

Voice Input: webkitSpeechRecognition (Web Speech API) for speech-to-text.

Voice Output: SpeechSynthesis for text-to-speech replies.

AI Engine: Gemini 2.0 Flash API via bolt.new (no-code/low-code setup).

Design: Bright colors, cartoon-style UI, clear buttons, large text for kids.

Hosting: Deployed on Netlify for easy access.

We created prompt templates for each mode to ensure the AI speaks in a safe, clear, and age-appropriate way.

Challenges we ran into

Making the TTS voice sound more natural instead of robotic.

Handling speech recognition for different accents and pronunciations.

Keeping response time short so conversations feel real and smooth.

Designing a UI that is both fun and easy for kids without overwhelming them.

Accomplishments that we're proud of

Built a fully functional, real-time AI tutor in a short hackathon timeframe.

Combined tutor-style learning with interactive roleplay in one app.

Created a safe, child-friendly experience with both fun and educational value.

Learned to integrate multiple APIs (speech-to-text, AI, TTS) smoothly.

What we learned

How to merge AI language models with live voice input/output.

The importance of UX design for very young users.

How prompt engineering changes AI tone and friendliness.

API optimization to ensure fast, low-latency AI responses.

What's next for GenieTalk

Add multi-language support for global reach.

Upgrade to more natural TTS voices (Google Wavenet, ElevenLabs).

Introduce AI memory to remember the child’s name, progress, and learning level.

Add learning progress tracking and parent dashboards.

Package as a PWA/Android app for offline and mobile use.

Built With

  • bolt.new
  • geminiapi
  • html
  • lowcode
  • tts
  • webtts
Share this project:

Updates